AI Briefingclip-vit-base-patch32: Lightweight version of CLIP for image classification without labels
openai/clip-vit-base-patch32
About the project
Classifies images not present in the learned class list using only text descriptions. This lightweight model adopts the ViT-B/32 architecture and performs zero-shot classification by calculating similarity between images and text.
It supports all major frameworks, including PyTorch, TensorFlow, and JAX. It can be easily integrated via AutoProcessor and AutoModel from the Transformers library, and is actively used with over 20 million monthly downloads.
This is a research model released by OpenAI to validate the generalization capabilities of computer vision. It is suitable for research and prototyping purposes rather than commercial deployment, and caution is required for non-English tasks as it is based on English datasets.
openai/clip-vit-base-patch32
The original page has no description.
zero-shot-image-classification
This introduction was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report errors, attribution issues, or removal requests via Contact.